py/objfloat: Support raising a negative number to a fractional power.

This returns a complex number, following CPython behaviour.  For ports that
don't have complex numbers enabled this will raise a ValueError which gives
a fail-safe for scripts that were written assuming complex numbers exist.

py/objfloat.c

@@ -298,6 +298,13 @@ mp_obj_t mp_obj_float_binary_op(mp_binary_op_t op, mp_float_t lhs_val, mp_obj_t
             if (lhs_val == 0 && rhs_val < 0) {
                 goto zero_division_error;
             }
+            if (lhs_val < 0 && rhs_val != MICROPY_FLOAT_C_FUN(floor)(rhs_val)) {
+                #if MICROPY_PY_BUILTINS_COMPLEX
+                return mp_obj_complex_binary_op(MP_BINARY_OP_POWER, lhs_val, 0, rhs_in);
+                #else
+                mp_raise_ValueError("complex values not supported");
+                #endif
+            }
             lhs_val = MICROPY_FLOAT_C_FUN(pow)(lhs_val, rhs_val);
             break;
         case MP_BINARY_OP_DIVMOD: {

tests/float/complex1.py

@@ -53,6 +53,10 @@ print(type(hash(1j)))
 # float on lhs should delegate to complex
 print(1.2 + 3j)
 
+# negative base and fractional power should create a complex
+ans = (-1) ** 2.3; print("%.5g %.5g" % (ans.real, ans.imag))
+ans = (-1.2) ** -3.4; print("%.5g %.5g" % (ans.real, ans.imag))
+
 # check printing of inf/nan
 print(float('nan') * 1j)
 print(float('inf') * (1 + 1j))